Research map / Adversarial robustness and security / Attacks on text and graphs
Imperceptible Adversarial Attack via Invertible Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use AdvINN-CGT instead of the strongest baseline | targeted attack generation · ImageNet-1K | LPIPS: 0.0352 -> 0.0118 | Helped |
| use AdvINN-HCT instead of SSAH | targeted attack generation · CIFAR-100 | LPIPS: 0.0200 -> 0.0035 | Helped |
| use AdvINN-CGT instead of PerC-AL | targeted attack generation · CIFAR-10 | LPIPS: 0.0134 -> 0.0030 | Helped |
| add IIEM to the generator (ablation) | AdvINN · ImageNet-1K | LPIPS: 0.0325 -> 0.0118 | Helped |
| add IIEM to the generator (ablation) | AdvINN · ImageNet-1K | iterations: 272 -> 321 | Mixed |
| replace IIEM with a CNN (ablation) | AdvINN · ImageNet-1K | LPIPS: 0.0118 -> 0.0360 | Hurt |
| replace IIEM with a CNN (ablation) | AdvINN · ImageNet-1K | FID: 1.594 -> 4.345 | Hurt |
| replace IIEM with a CNN (ablation) | AdvINN · ImageNet-1K | iterations: 321 -> 1800 | Hurt |
| use a learnable classifier-guided target image instead of a fixed target image | AdvINN · ImageNet-1K | LPIPS: 0.0206 -> 0.0118 | Helped |
| use UAP instead of HCT as the target image | AdvINN · ImageNet-1K | optimization speed: around 20% speed-up | Helped |
| learn a classifier-guided target image instead of using UAP | AdvINN · ImageNet-1K | iterations to converge: only takes around 10% iterations | Helped |
| vary the adversarial budget among 4/255, 8/255, and 16/255 (ablation) | AdvINN · ImageNet-1K | ASR (%): 100.0; 100.0; 100.0 | No clear effect |
Closest papers by meaning
- Towards Imperceptible and Robust Adversarial Example Attacks Against Neural Networks (AAAI 2018 · similarity 0.89)
- Constructing Unrestricted Adversarial Examples with Generative Models (NeurIPS 2018 · similarity 0.88)
- Evading Defenses to Transferable Adversarial Examples by Translation-Invariant Attacks (CVPR 2019 · similarity 0.88)
- A Direct Approach to Robust Deep Learning Using Adversarial Networks (ICLR 2019 · similarity 0.88)
- One Man's Trash Is Another Man's Treasure: Resisting Adversarial Examples by Adversarial Examples (CVPR 2020 · similarity 0.87)
- AdvAD: Exploring Non-Parametric Diffusion for Imperceptible Adversarial Attacks (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.